Assessing Municipal Energy Intensity Based On Nighttime Light Data Of China

The excessive consumption of energy emits a large amount of carbon dioxide,which affects the balance of earth-atmosphere radiation,intensifies the greenhouse effect,and leads to global warming.The macro carbon neutralization goal proposed by China requires strict control and improvement of the current energy utilization efficiency.Energy intensity is an important indicator of energy utilization efficiency.Research on energy intensity at a fine scale can help to solve the problem of excessive energy consumption in the region,thus promoting overall energy conservation and emission reduction.However,only provincial and above energy statistics are officially released,which hinders the study of energy intensity on a fine scale and the formulation of precise energy economic policies.Therefore,it is necessary to estimate and analyze energy intensity at municipal level.The earth observation data represented by nighttime light can effectively describe the intensity of human activities and thus provide the possibility for estimating municipal energy intensity.However,due to the lack of inter-annual consistency in the night light data collected by multi-source sensors,the research of long-term energy intensity was hindered.In order to realize the continuous estimation of municipal energy intensity for many years,this study first calibrate the multi-sensor nighttime light data to generate a continuous long-term product of nighttime light.Then the product and other earth observation data are used to estimate the municipal energy intensity by machine learning algorithm.Finally,the temporal and spatial analysis are carried out to provide targeted policy recommendations.It is expected that this study can support the completion of the overall sustainable development goals.An inter-calibration algorithm is developed in this study to integrate the nighttime light data of DMSP-OLS and NPP-VIIRS,which lays a foundation for the study of energy intensity of long-term series.Firstly,the region with stable and uniform light radiation is selected as the vicarious site within the whole country.By comparing different models,Bi Dose Resp is selected as the key function to realize the preliminary calibration of NPPVIIRS,and its coefficient of determination reaches 0.967.Then the Gaussian Low-Pass Filter with the optimal parameters is used to smooth the result of preliminary calibration to maximize the consistency between the two sensors.The results show that the correlation coefficient between the images of DMSP-OLS and the calibrated NPP-VIIRS at the resolution of 1000 m within the whole country reaches 0.949 in 2013,whereas before calibration,the correlation coefficient was only 0.621.By comparing the image texture,light profile and pixel relationship between DMSP-OLS and the calibrated NPP-VIIRS from both macro and micro perspectives,it is found that they are highly consistent.Finally,this algorithm is used to produce the continuous nighttime light data at the kilometer level from1992 to 2018 within whole country.This product can promote long-term data mining and provide effective support for the study of human activities.Based on the calibrated nighttime light data and other Earth observation data,machine learning algorithms are used to estimate energy intensity at the municipal level.Features are extracted from multi-source data through feature engineering,and then cross-validation is used to evaluate the ability of multiple linear regression model,random forest model and integrated deep neural network model to predict energy intensity.During this process,the appropriate methods of adjusting hyperparameter are adopted for different models to improve the capability of the models.Finally,it is found that the average and RMSE on test set of the integrated deep neural network model are 0.913 0.004 and 215.93318.327 respectively.Compared with other models,the integrated deep neural network model has the highest and the most stable estimation ability.Therefore,this model is selected to estimate the municipal energy intensity of the mainland China from 2001 to 2017.Finally,a temporal and spatial analysis of municipal energy intensity is conducted to discuss the development of municipal energy intensity during this period.First,by observing the overall spatial distribution of municipal energy intensity,it is found that the energy intensity gradually increases from the southeast to the northwest,indicating that the energy efficiency of cities in Southeast China is higher than that in Southwest China.By observing the changes of the peak and mean of the energy intensity histogram at the municipal level over time,it is discovered that,from 2001 to 2017,most cities in the Chinese mainland have been developing towards the direction of improving energy efficiency.Through linear fitting of energy intensity for each city,the slope is taken as the change speed of energy intensity.It is found that the cities in central region have the fastest decline rate,followed by those in eastern region,and the cities in western region have the slowest decline rate.The spatial autocorrelation analysis about the decline rate of energy intensity shows that the decline rate of energy intensity presents obvious spatial aggregation,which indicates that energy policy has a multiple-block-joint effect among cities.Finally,through the Initial Energy Intensity-Slope Analysis,it is found that the energy intensity of a few cities during this period did not decrease but increased.The reasons for the increase are different,which is worth attention of policy makers.In addition,some cities with extremely rapid decline in energy intensity and those with very low energy intensity for a long time are found,and they can play an exemplary role in the green transformation.The assessment and analysis of urban energy intensity offers a new perspective and new method to study energy efficiency for the academic community,and provides effective information to accurately manage energy efficiency issues for policy makers.It is expected that this study can promote China to achieve energy conservation and emission reduction targets.